Control and Optimization Meet the Smart Power Grid - Scheduling of Power Demands for Optimal Energy Management

Abstract

The smart power grid aims at harnessing information and communication technologies to enhance reliability and enforce sensible use of energy. Its realization is geared by the fundamental goal of effective management of demand load. In this work, we envision a scenario with real-time communication between the operator and consumers. The grid operator controller receives requests for power demands from consumers, with different power requirement, duration, and a deadline by which it is to be completed. The objective is to devise a power demand task scheduling policy that minimizes the grid operational cost over a time horizon. The operational cost is a convex function of instantaneous power consumption and reflects the fact that each additional unit of power needed to serve demands is more expensive as demand load increases.First, we study the off-line demand scheduling problem, where parameters are fixed and known. Next, we devise a stochastic model for the case when demands are generated continually and scheduling decisions are taken online and focus on long-term average cost. We present two instances of power consumption control based on observing current consumption. First, the controller may choose to serve a new demand request upon arrival or to postpone it to the end of its deadline. Second, the additional option exists to activate one of the postponed demands when an active demand terminates. For both instances, the optimal policies are threshold based. We derive a lower performance bound over all policies, which is asymptotically tight as deadlines increase. We propose the Controlled Release threshold policy and prove it is asymptotically optimal. The policy activates a new demand request if the current power consumption is less than a threshold, otherwise it is queued. Queued demands are scheduled when their deadline expires or when the consumption drops below the threshold.